> I'm not sure that Daniel's second is recognized by Robert's Rules of Order, > but I am adding a similar test to the docs > > from fipy import parallel, Grid1D > > mesh = Grid1D(nx=10) > print "%d cells on processor %d of %d" % (mesh.getNumberOfCells(), > parallel.procID, parallel.Nproc) > > that confirms both that multiple processors are in play and that FiPy is > actually making use of them.
An important feature of the Wheeler Test, from my perspective, is that it probes just the FiPy infrastructure, depending only on the trilinos/MPI build. For completeness, a similarly simple bit of code expressing the same idea using only mpi4py/MPI can be imagined [from mpi4py import MPI .. tot_procs= MPI.COMM_WORLD.Get_size() rank= MPI.COMM_WORLD.Get_rank() ..] In combination with these two tests, the Guyer Test (using parallel & Grid or some- thing like it), would give the novice some confidence that all the individual com- ponents were functioning properly, or at least provide some help in narrowing the scope of the problem. +jtg+
